Closed johnynek closed 6 years ago
@ttim @erik-stripe @non ptal
note we ran these rules in production on storm and significantly reduced the number of storm nodes we created.
I think you should consider using these to optimize your storm graphs as well.
Yeah I can second this recommendation. 👍
These tests are in a very bad place.
I have been clicking rebuild to deal with storm or hadoop timeouts for about 24 hours now. Each trial is another 20-60 minutes. I am almost ready to merge without the tests...
We need to really think about improving the platform tests.
Merging #758 into develop will decrease coverage by
1.07%
. The diff coverage is62.82%
.
@@ Coverage Diff @@
## develop #758 +/- ##
===========================================
- Coverage 72.5% 71.43% -1.08%
===========================================
Files 154 151 -3
Lines 3783 3606 -177
Branches 211 209 -2
===========================================
- Hits 2743 2576 -167
+ Misses 1040 1030 -10
Impacted Files | Coverage Δ | |
---|---|---|
.../scala/com/twitter/summingbird/memory/Memory.scala | 94.73% <ø> (ø) |
:arrow_up: |
.../twitter/summingbird/memory/ConcurrentMemory.scala | 95.77% <100%> (+0.06%) |
:arrow_up: |
...com/twitter/summingbird/planner/DagOptimizer.scala | 60% <62.33%> (-14.02%) |
:arrow_down: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 57312f7...f78f056. Read the comment docs.
Scalding 0.18 is using dagon to implement its new backend. We might as well use dagon here too, which was taken out from summingbird about 6 months or so ago.
We have found a number of performance improvements in the code since then, so it should be a win.